In the conventional image coding standards represented by International Telecommunication Union Telecommunication Standardization Sector (ITU-T) standards denoted with H. 26x and International Organization for Standardization/International Electrotechnical Commission (ISO/IEC) standards denoted with MPEG-x, a picture is partitioned into predetermined units and coding is performed on each of the partitioned units. For example, in the H. 264/MPEG-4 AVC standard (for example, refer to Non Patent Literature 1), a screen (picture) is processed by a unit of horizontal 16 pixels and vertical 16 pixels, which is referred to as a macroblock. When motion compensation is performed, data can be compressed after a macroblock is partitioned into blocks (each of the blocks is composed of a minimum unit of horizontal 4 pixels×vertical 4 pixels), motion compensation is performed on each of the blocks with a different motion vector, frequency transform is performed on a difference signal that is a difference with an original signal, the difference signal is collected into a low frequency range and quantization is performed on the difference signals. Generally, a transform block size having a large size is advantageous because correlation can be more effectively used.